How to Improve Your Knowledge Graph

How to Improve Your Knowledge Graph

Option 1: Chat to Explore

Ask questions to understand what Aether found:

"What entities did you find in the sales data?"

"Show me all the concepts you discovered"

"Why is deal SAAS-3 marked as at_risk?"

"What rules have the highest confidence scores?"

Option 2: Chat to Refine

Add your domain expertise:

"Create a concept called 'vip_customer' for deals over $500k"

"Add 'Territory' as a new entity type"

"Update the relationship between Sales Rep and Manager to 'reports_to'"

"Remove the duplicate entity for ABC Company"

The system learns from your refinements and applies them across your entire project.

Option 3: Run Deeper Analysis

If your initial coverage is low, request additional pattern discovery:

"Run correlation analysis on deal velocity"

"Find clustering patterns in customer behavior"

"Analyze what predicts deal success"

Each analysis can add 5-10 new concepts and rules.

Tips

• Upload comprehensive data first: More data yields better pattern discovery (aim for 1,000+ rows minimum)
• Review coverage after upload: If below 20%, your data may be too sparse or require custom concepts
• Let Aether run both processes: Don't interrupt—structure detection and pattern discovery happen simultaneously
• Start with factual queries: Test basic relationships before complex analysis
• Add your expertise gradually: Let the system show you what it found before overriding with custom concepts
• Domain matters: People-centric workflows (Sales, HR) and process-centric workflows (Supply Chain, Manufacturing) create different graph structures—both work perfectly

What Happens Behind the Scenes

When you upload data, Ontology Builder creates the skeleton:

• Detects entity types from column patterns
• Infers dimensions from data structure
• Identifies workflow stages
• Maps basic relationships